We investigated the impact of an HH intervention for nursing homes staff on HAI in residents. Five illnesses were investigated: gastroenteritis, ILI, pneumonia, UTI, and MRSA. There was statistically significantly more gastroenteritis (p<0.001) and less ILI (p<0.01) in the intervention arm when compared to the control arm when taking baseline data into account and controlling for the clustering of observations in nursing homes, baseline differences, and the period in the study. For pneumonia and UTI, there were no differences between study arms. Sensitivity analysis did not confirm that there was statistically significantly more gastroenteritis in the intervention arm (p=0.92).
Other studies have also looked at the effect of an HH intervention on HAI reduction. The results of the systematic review cited in the introduction suggest that HH interventions may help control the infectious risk in nursing home settings, but that the precise impact remains poorly documented [6]. Many studies in the review were limited by methodological flaws; only 8 of 56 studies were RCTs, 6 of which were published over the last 5 years before the review (in the period 2010–2015). Also, most studies were in single-site nursing homes and provided a limited array of data. Finally, a low proportion of the studies in the review included direct observations of HH compliance, and the authors recommend strongly that future studies should include direct observation of HH compliance. Our current study complies with the recommendations from this review in that it is a large multicenter trial with extensive data collection on many possible determinants for HH compliance and risk factors for infection. Additionally, HH compliance was established through direct observation. Nevertheless, our study produced rather paradoxical results of which the interpretation is challenging.
After baseline, nursing homes were randomly assigned to either the control or intervention arm, ensuring that nursing homes from the same organization were in different study arms. Despite randomization, there were marked differences in the distribution of HAIs at baseline. This could possibly be explained by the fact that the introduction of infectious disease is a highly random phenomenon, especially when observed over a short period. Baseline differences between the two arms of the trial were particularly notable for ILIs and to lesser extent for gastroenteritis.
Many infectious diseases are seasonal. We addressed this through our RCT-design, assuming the seasonal changes to be the same in both arms. Yet, our implementation of the RCT may not have been ideal for two reasons: (1) It was possibly difficult to see a statistically significant difference between the study arms, because the follow-up period was primarily after the winter season when one would expect lower rates of gastroenteritis and ILI (February–October 2017); and (2) because of the generally low HAI incidence, observation is ideally performed over multiple years.
A hand hygiene intervention is not always the most important hygiene intervention to reduce HAIs, which can have both endogenous and exogenous sources. Hand hygiene compliance should primarily decrease HAIs that spread through person-to-person contact, with a secondary effect of lower contamination of surfaces. When the most prevalent transmission route is via droplet or aerosols, mask usage can be the most important hygiene intervention. We would therefore expect the effect of increased hand hygiene on gastroenteritis to be high and on pneumonia or UTI to be low, considering the disease pathways. At the same time, hand hygiene is necessary when handling a catheter and approximately 12% of nursing home residents have a catheter [13]. The results of our study are rather paradoxical (there was a statistically significant increase of gastroenteritis in the intervention arm) and emphasize that it is difficult to establish the effect of improved hand hygiene when using HAI as an outcome indicator.
To place the outcomes of the HANDSOME study into perspective, we compared these with a Dutch national surveillance program (SNIV) and European data from the ECDC. The nursing homes in HANDSOME (both intervention and control arms of the study) followed the SNIV data closely (except for ILIs); the control arm followed the SNIV trends more closely. A possible explanation that the nursing homes in the intervention arm registered more infections could be that the nursing homes in the intervention arm were extra alert to infections among residents because of the intervention and thus more motivated to provide diligent illness incident reports than nursing homes in the control arm. Comparing our data to the infection rates provided by the ECDC, we had slightly more reporting of HAI (4.2 per 1000 resident days vs. 3.2 per 1000 resident days), even though the ECDC has a broader definition of HAI, including, for instance, skin/soft tissue infections, eye/ear/mouth/nose infections and bloodstream infections [1].
We used the McGeer criteria in this study to define infectious diseases for two reasons: (1) the national SNIV uses the McGeer criteria, and we wanted to compare our data to another dataset; and (2) it is hard to justify (invasive) diagnostic testing in nursing home residents when the goal of the study is not to find suspected HAIs but to understand the effect of hand hygiene on HAIs. At the same time, the diagnosis of HAIs is often uncertain and may be based on subjective criteria. Additionally, the McGeer criteria have been updated by diverse researchers and organizations; newer insights could lead towards more accurate identification of HAIs [14, 15]. Future studies could perform diagnostics for more definitive results or use updated versions of the McGeer criteria.
The effect of HH on HAIs may be dependent on various infection prevention measures, such as cleaning methods and schedules. It is also assumedly dependent on the HH compliance level. Although the HANDSOME intervention was successful in tripling the HH compliance in the intervention arm, it only reached a 36% compliance rate [7]. The hand hygiene compliance in the intervention arm may not have crossed a critical threshold to lower infection rates. Some (primarily single-site or small-scale) studies in nursing homes have shown a correlation between HH compliance and infection rates, although larger studies generally show no relationship, making it difficult to determine a threshold value [6]. The compliance rate after the intervention might have been higher if more nurses had attended the lessons; the estimated attendance of health care workers at at least one of the lessons varied per unit: 23% units had <50% of the unit’s health care workers attending at least one lesson, 18% had 50–74% attendance at at least one lesson and 59% had >75% attendance at at least one lesson (n=22).
Understanding the pathways of HAIs during social interactions in nursing homes is also important when evaluating the results of interventions on HAIs. In contrast to hospital settings, nursing homes promote the socialization of residents. Residents may practice poor hygiene, and hence infect each other. The HANDSOME intervention did not target residents. Therefore, it cannot be expected that the direct resident-to-resident infection rate decreased. There are also social interactions in a nursing home between residents and staff for which HH is not prescribed by the WHO, such as a handshake or patting a hand [16]. This is different than in a hospital, where all hand interactions are considered HH opportunities [17].
A strength of the study is that it is based on data from a large multicenter cluster RCT. There are also limitations. Firstly, there may have been factors that influenced the reliability of the HAI data. Illness was recorded weekly by hand, which could elicit recall bias. Although the nursing home staff was accustomed to reporting infections in individual dossiers, they were not accustomed to reporting weekly infections for the unit as a whole. Since this type of illness incident reporting was new, it may have taken time until the illness incident reporting was consistent. Secondly, consistency between units may also have been a problem, since the function of the staff member who registered illnesses (nurse, team manager, or geriatrician) varied per unit. The staff member’s knowledge of HAIs present in the unit may also have differed. At the same time, this study used stratified randomization; for every nursing home in the intervention arm (2 units), there was generally one nursing home from the same organization in the control arm (2 units), thereby minimizing differences between study arms. The nursing homes in the two study arms were also not statistically significantly different for various variables, including management style, number of nurses per resident, and the intensity of care [7]. Therefore, we expect the illness incident reporting errors to be similar in the two arms of the study.